高頻率財務資料的隨機波動性之估計方法探討

Abstract

本論文主要在於探討不同方法對於波動率在高頻率資料中的估計效果差異。 在統計及財務領域，估計股價的隨機波動率是很重要的課題，因此我們首先介 紹對低維度財務資料基本的隨機差分模型以及直觀解決高頻率問題的 TSRV 估 計式，接著面對維度資料時，引進 Wishart Autoregressive process 及 Combination Approaches 二種波動率估計方法，最後使用模擬的方式比較兩種 方法之間的差異。

The thesis focus on the analysis of high-frequency financial data. There are lots of estimation methods in univariate asset such as two-time scale realized volatility (TSRV) which we would introduced here. In multiple assets, we present the Wishart-Autoregressive process which can be used to model the dynamic structure of volatility matrices in financial application, with the closed-form prediction and model flexibility, it is alternative to GARCH or stochastic Models. Another method is combination approaches which can simply compute the estimations or predictions without losing any information of available data. Finally, we compare the estimations for intraday returns in stochastic volatility.